Governance through Language: The Thesaurus as a Management Tool
++++++++++++++++++++++++++++++++++++++++++++++++
SHORT ABSTRACT
Taxonomies, thesauri and dictionaries have a long history as a means to capture knowledge
and the language by which it is expressed. In organizational contexts, there is a growing
interest in creating such artefacts. By doing so, organizations strive to attain a variety of
related goals: capturing and dissemination of knowledge, normalization of terminology, and
quality assurance — to mention a few. Increasingly, thesauri play a role in governing the
design and operation of IT-systems. This is an interesting development, as it shifts the control
over what data mean from the IT Department towards the business.
The shift results from a number of trends. Technology providers respond to these in
innovative ways. Standards-compliant thesaurus management software offers an increasing
number of features. This includes automatic “seeding” of a thesaurus by mining the DBPedia
or large text corpora, but also mechanisms to govern “pick lists” such as “gender: {male,
female, undisclosed}”, so that software applications used in business processes literally use
the thesaurus as the single source of reference data. An exciting development is adding
features that leverage semantic integration through Linked Data technologies.
We will present an analysis of these trends. We also reflect on the different strategies to deal
with polysemy, idiosyncrasies, and multilingualism and to organize governance. Finally, we
discuss a real-world case.
++++++++++++++++++++++++++++++++++++++++++++++++
DESCRIPTION
The problem addressed in this presentation is ultimately a problem of governance.
As IT systems become larger, more complex and more important, assuring the quality of the
information inside them starts to become an issue. Managing definition of terms through data
dictionaries containing tens or hundreds of thousands of data items simply doesn’t work.
Moreover, it is the IT-department that, often unwillingly, is responsible for ownership. The
alternative is a thematically organized thesaurus with a manageable amount of concepts, with
broadly used and standardized methods to capture semantic relations between terms, and to
deal with multilingualism, synonymy and homonymy. A group of executives within the
business can then be formed to act as owner and regain control over what the information
inside the IT-systems means. This is a prerequisite for managing data quality.
A second trend is the push for large scale, cross-border integration of IT systems, or rather, of
the information contained in them. Technological interoperability standards, such a SOAP
and REST, have enabled this push. At the same time, however, interoperability at the
semantic level is a matter of human language. The same sociolinguistic principles that govern
human interaction are at work. This means, for instance, that forcing shared data dictionaries
on many organizations in a network in a top-down manner will meet impedance. Yet, this is
the classic paradigm for creating interoperability standards. Modern approaches have learned
from the rise of the Web: bottom up, allowing for mismatches by defining mechanisms to
deal with these. What is needed, however, is allowance for “pidginization” and “creolization”
— that is, a language that develops in interaction rather than one being designed by an
authority.
The innovative nature of this approach is the application of long-known techniques and
methods from the library sciences in the domain of IT-governance. This application poses
new challenges which can be elegantly met using Linked Data techniques.
Technology providers respond with creative solutions. Standards-compliant thesaurus
management software offers an increasing number of features beyond what is needed for the
traditional application of indexing libraries. This includes relatively simple but highly
effective mechanisms to govern “pick lists” such as “gender: {male, female, undisclosed}”,
so that software applications used in business processes literally use the thesaurus as the
single source of reference data. An exciting development is adding features that leverage
semantic integration through Linked Data technologies for instance for the purpose of
automatic “seeding” a thesaurus by mining the DBPedia or large text corpora.
A recent example of the latter development is an open thesaurus for legal terms published by
Wolters Kluwer Germany (http://vocabulary.wolterskluwer.de), which is a controlled
vocabulary enriched with information from DBPedia and Eurovoc.
The so-called “justitiethesaurus” is an example of an important, authoritative thesaurus in the
justice domain in the Netherlands. Originally developed for the purpose of information
retrieval in the research library published by the WODC, the research bureau of the
Department of Security and Justice, it is currently used in modified and enriched form by
several other organizations for their libraries. Increasingly, the thesaurus is looked at with an
eye to use it as a basis for term management in IT-systems.
In the world of library management, the governance is often well-organized. The example of
the “justitiethesaurus” shows how different organizations can cooperate effectively and reuse
parts of each other’s vocabulary, with enough flexibility to avoid paralysis. In the world of
structured data, this type of cooperation is still perceived as challenging.
The central theme of this presentation is governance over the vocabulary through which the
operation of an organization is manifested. Based on several examples, lessons learned will
be drawn concerning strategies to organize this type of governance. On the one hand, a
thesaurus must be authoritative so that it offers some measure of stability and quality. On the
other, this authority must be localized enough so as to allow for effective and efficient
decision making.
++++++++++++++++++++++++++++++++++++++++++++++++
LONG ABSTRACT
Taxonomies, thesauri and dictionaries have a long history as a means to capture knowledge
and the language by which it is expressed. In organizational contexts, there is a growing
interest in creating such artefacts. By doing so, organizations strive to attain a variety of
related goals: capturing and dissemination of knowledge, normalization of terminology, and
quality assurance — to mention a few. Increasingly, thesauri play a role in governing the
design and operation of IT-systems. This is an interesting development, as it shifts the control
over what data mean from the IT Department towards the business.
The shift results from a number of trends. As IT systems become larger, more complex and
more important, assuring the quality of the information inside them starts to become an issue.
Managing definition of terms through data dictionaries containing tens or hundreds of
thousands of data items simply doesn’t work. The alternative is a thematically organized
thesaurus with a manageable amount of concepts, with broadly used and standardized
methods to capture semantic relations between terms, and to deal with multilingualism,
synonymy and homonymy.
A second trend is the push for large scale, cross-border integration of IT systems, or rather, of
the information contained in them. Technological interoperability standards, such a SOAP
and REST, have enabled this push. At the same time, however, interoperability at the
semantic level is a matter of human language. The same sociolinguistic principles that govern
human interaction are at work. This means, for instance, that forcing shared data dictionaries
on many organizations in a network in a top-down manner will meet impedance. Yet, this is
the classic paradigm for enabling data exchange. Modern approaches have learned from the
rise of the Web: bottom up, allowing for mismatches by defining mechanisms to deal with
these. What is needed, however, is allowance for “pidginization” and “creolization” — that
is, a language that develops in interaction rather than one being designed by an authority.
Technology providers respond to these trends in innovative ways. Standards-compliant
thesaurus management software offers an increasing number of features. This includes
automatic “seeding” of a thesaurus by mining the DBPedia or large text corpora, but also
relatively simple but highly effective mechanisms to govern “pick lists” such as “gender:
{male, female, undisclosed}”, so that software applications used in business processes
literally use the thesaurus as the single source of reference data. An exciting development is
adding features that leverage semantic integration through Linked Data technologies.
A recent example of the latter development is an open thesaurus for legal terms published by
Wolters Kluwer Germany (http://vocabulary.wolterskluwer.de), which is a controlled
vocabulary enriched with information from DBPedia and Eurovoc.
We will present an analysis of these and related trends. We also reflect on the different
strategies to deal with polysemy, idiosyncrasies, and multilingualism. Finally, we discuss
ways to organize the processes for creating and managing the vocabulary and their
governance. Finally, we discuss a real-world case.